Method and system for training image-alignment procedures in computing environment

Abstract

The present subject matter refers a method for training image-alignment procedures in a computing environment. The method comprises communicating one or more images of an object to a user and receiving a plurality of user-selected zones within said one or more through a user-interface. An augmented data-set is generated based on said one or more images comprising the user-selected zones, wherein such augmented data set comprises a plurality of additional images defining variants of said one or more communicated images. Thereafter, a machine-learning based image alignment is trained based on at-least one of the augmented data set and the communicated images.

What is claimed is: 
     
       1. A method for training image-alignment procedures in a computing environment, said method comprising:
 communicating one or more images of an object to a user; 
 receiving a plurality of user-selected zones within said one or more through a user-interface; 
 generating an augmented data set based on said one or more images comprising the user-selected zones, said augmented data set comprising a plurality of additional images defining variants of said one or more communicated images; and 
 training a machine-learning based image alignment method based on at-least one of the augmented data set and the communicated images. 
 
     
     
       2. The method as claimed in  claim 1 , wherein the plurality of user selected-zones are defined by at least one of:
 i. a plurality of point-selections within the communicated images; and 
 ii. a plurality of portions selected by a user within the communicated images. 
 
     
     
       3. The method as claimed in  claim 1 , further comprising:
 identifying the object within the one or more communicated images based on the user-selected zone; 
 subjecting the identified object within one or more communicated images to at least one of: rotation, inversion, scaling, displacement; and 
 generating the variants of the one or more communicated images forming a part of the augmented data set based on said subjection to the identified object. 
 
     
     
       4. The method as claimed in  claim 1 , wherein the communication of the one or more images to the user comprises shortlisting the image having a substantially less distortion to thereby facilitate an ease of selection of the zones by the user through the user interface. 
     
     
       5. The method as claimed in  claim 1 , wherein said communicated one or more images and the augmented data set define at least one of:
 a training data set; 
 a validation data set; and 
 a testing data set. 
 
     
     
       6. The method as claimed in  claim 1 , further comprising:
 executing the machine-learning based image alignment method with respect to real-time image data; and 
 communicating the aligned images with respect to the real-time image data to an image-inspection procedure as a part of image quality-control process. 
 
     
     
       7. The method as claimed in  claim 1 , wherein the plurality of user selected-zones within the communicated images are defined by at least one of:
 a plurality of corners of the object; 
 a plurality of edges of the object; 
 one or more boundaries of the object; and 
 a free form shape drawn by the user within the communicated images to localize the object. 
 
     
     
       8. The method as claimed in  claim 7 , wherein the plurality of user selected-zones within the communicated images correspond to a set of common-features of the object across the plurality of images. 
     
     
       9. The method as claimed in  claim 1 , wherein the machine learning (ML) based image alignment method comprises operating upon a generic ML image alignment procedure through one or more of:
 undergoing training through a training data set as a part of training phase; 
 detecting an object within an real-time image data set; and 
 modifying a position of the detected object with respect a frame within the real-time image data set to thereby align the object within the frame in accordance with a pre-defined standard. 
 
     
     
       10. The method as claimed in  claim 9 , wherein said machine learning based image alignment method is at least one of:
 a deep learning procedure and 
 a convolution neural network. 
 
     
     
       11. A method for training image-alignment procedures for facilitating image-inspection, comprising:
 communicating one or more images of an object to a user; 
 receiving a plurality of user-selected zones within said one or more images through a user-interface; 
 generating an augmented data set based on said one or more image comprising the user-selected zones, said augmented data set comprising a plurality of additional images defining variants of said one or more communicated images; 
 training a machine-learning based image alignment method based on at least one of the augmented data set and the communicated images; and 
 executing the machine-learning based image alignment method with respect to a real-time image data and communicating aligned images for image inspection. 
 
     
     
       12. The method as claimed in  claim 1 , wherein the plurality of user selected-zones are defined by two or more point-selections executed by the user within the communicated images. 
     
     
       13. The method as claimed in  claim 1 , further comprising:
 subjecting the object within one or more communicated images to automatically undergo one or more image editing techniques defined by at least one of: rotation, inversion, scaling, displacement; and 
 generating the variants of the one or more communicated images forming a part of the augmented data set based on said image-editing techniques. 
 
     
     
       14. The method as claimed in  claim 1 , wherein the communication of the one or more images to the user comprises shortlisting the image having a substantially less distortion to thereby facilitate an ease of selection of the zones by the user through the user-interface, said shortlisting defined by one or more of:
 a user performed shortlisting through manual action; and 
 an execution of dendrogram to cause automatic shortlisting. 
 
     
     
       15. The method as claimed in  claim 1 , wherein the plurality of user selected-zones within the communicated images are defined by:
 a plurality of corners or edges in case of a polygonal object; 
 a free-form boundary drawn by a user around the object exhibiting a circular or elliptical shape. 
 
     
     
       16. The method as claimed in  claim 15 , wherein the plurality of user selected-zones within the communicated images correspond to a set of common-features of the object across the plurality of images. 
     
     
       17. The method as claimed in  claim 1 , further comprising:
 obtaining one or more aligned images from an operation of the machine-learning based image alignment procedure upon a real-time image data set of an object; 
 communicating the aligned-images to an image-inspection method for enabling a quality-control of the real-time image data. 
 
     
     
       18. The method as claimed in  claim 17 , wherein the image-inspection method is a machine learning method for certifying an image of an object as acceptable, permissible, unacceptable, prone to be rejected, prone to be accepted. 
     
     
       19. A non-transitory medium comprising computer-executable instructions which, when performed by processor cause the processor to train image-alignment procedures for facilitating image-inspection by the steps of:
 communicating one or more images of an object to a user; 
 receiving a plurality of user-selected zones within said one or more images through a user-interface; 
 generating an augmented data set based on said one or more image comprising the user-selected zones, said augmented data set comprising a plurality of additional images defining variants of said one or more communicated images; 
 training a machine-learning based image alignment method based on at least one of the augmented data set and the communicated images; and 
 executing the machine-learning based image alignment method with respect to real-time image data and communicating aligned images to an image-inspection procedure.